Problems of validity and reliability of concept mapping are addressed by using item-response theory (IRT) models for scoring. In this study, the overall structure of students' concept maps are defined by the number of links, the number of hierarchies, the number of cross-links, and the number of examples. The study was conducted with 92 students in four classes at a junior high school in Canada. Results show that IRT scoring of concept maps is generally valid and reliable. The correlation between IRT ability estimates and the total concept-mapping scores based on a scoring scheme proposed by J. D. Novak is significant, demonstrating that it is valid to score concept maps on the basis of structural characteristics defined by links, hierarchies, cross-links, and examples. The advantage of IRT scoring is reliability. Some computer packages with concept-mapping facilities and IRT scoring are described. One figure and eight tables present study findings. (Contains 29 references.) (SLD)